There are different data sets available with ML repositories. It is good to check the cite..

Here you are:

http://archive.ics.uci.edu/ml/index.php

Have you thought of using Twitter for sentiment analysis? There is labeled data available, for instance here:

http://help.sentiment140.com/for-students

Yes I agree with Peter, the cite provide not only labeled data but also a description of algorithm used and ML approach they applied, and more importantly a link to open source codes for Twitter sentiment analysis using Python and NLTK.